1. Introduction

Blood leakage at the site of endothelium failure is counteracted by a two-step process, the haemostasis: the primary haemostasis step involves the thrombocytes (platelets) which come together to form a plug, while in the secondary haemostasis step, the coagulation factors of the bloodstream form a fibrin meshwork [1]. Platelets are 2–3 µm biconvex disc exclusively found in the blood of mammals. They form a plug in a stepwise mechanism. First, circulating platelets are recruited at the site of injury by the exposure of deep tissue structures. The adhesion of these platelets to the surface triggers the recruitment and aggregation of novel platelets. Upon platelet adhesion, surface receptors are activated [2]. Thanks to these receptors, platelets are bridged together through the interaction with factors such as the fibrinogen or the Von Willebrand factor (vWF) that harbor multiple binding sites [3]. Activation also triggers morphological changes of the platelets. Upon adhesion on the wound, the platelets flatten on the surface and develop filopodia which intertwine to tighten the platelet aggregate [4]. For more than 4 decades, platelet aggregation was depicted following a simple unified model requiring a stimulus, which was a soluble protein (fibrinogen), and a membrane-bound platelet receptor (integrin αIIbβ3 or GPIIb-IIIa) [5]. However, recent technical advances allowing real time analysis of platelet aggregation in vitro and in animal models demonstrated much more complex dynamical processes than previously expected [6]. In particular, the mechanisms by which hemodynamic conditions lead to platelets adhesion and aggregation are still incompletely understood [6,7]. Actual results suggest that platelet tethering requires different receptor/ligand pairs at low (up to 1000 s⁻¹: fibrinogen/integrin αIIbβ3) and high (up to 10000 s⁻¹: vWF/ GPIb glycoprotein bonds) shear rates of the bloodstream [8].

In clinical practice, platelet functional tests read out exhaustive information about the role of the different receptors present at the surface of platelets [9]. These tests are based on turbidimetric optical detection, multiple electrode aggregometry and flow cytometry [10]. Using such tests, it is impossible to analyze the processes that follow platelets adhesion, namely the spreading and the retraction although they both play a pivotal role in the platelet adhesion and the thrombus formation [11].

To our knowledge, quantitative 3D morphology of platelet aggregation and spreading has never been studied due to technical limitations. However, these limitations can be overcome by using holographic technology. Indeed, the Digital Holographic Microscopy (DHM) is a technology that allows quantitative phase contrast imaging and digital refocusing from a single recorded hologram [12–14]. Recently, the DHM has been used to determine the 3D morphometry of human Red Blood Cells (RBC) [15,16]. 3D platelet aggregates analysis is an important challenge to better understand the spreading mechanisms in hemostasis physiology and in pathology. In this work, we develop a novel method able to characterize the 3D aggregate characteristics such as aggregates surface, height and volume by using the quantitative phase contrast imaging capability of DHM. In order to reduce as far as possible the influence of defects and aberrations, we use a correction strategy similar to the one described in [14].

2. Materials and methods

Aggregates formation: the impact-R test

Blood sampling was approved by the CHU Charleroi hospital ethics committee (Comite´ d’Ethique I.S.P.PC: OM008). The studies conform to the principles outlined in the Declaration of Helsinki. Venous blood was drawn from 3 healthy donors (from the Centre Hospitalier Universitaire de Charleroi, Belgium) into tubes with 3.2% sodium citrate solution, pH 7.4. A cone and plate device (Impact-R, Diamed) was used [17] (Fig. 4(a)). The platelet aggregates formation was induced by exposing 130 µL of whole blood in a well to laminar flow using the disposable Teflon conical rotors. After washing, images on a circumferential plane from the wells were captured by the image analyzer on impact-R, which quantifies the platelet aggregates formed, in 2D, on the surface. The results are expressed as number of aggregates detected and the average size of the aggregates, according to the shear stress. This device can produce laminar flows in the range of 25 s⁻¹ to 5000 s⁻¹ and with exposition times between 20 and up to 300 seconds maximum. For the present study, the 2 edges from the exposition times were analysed and compared. After testing different shear rates (25, 100, 400, 1800, 5000) we selected the one of 100 s⁻¹ since it corresponds to the experimental condition where more platelet aggregates are created on the polystyrene surface of the Impact-R test well (data not shown).

The Fig. 1 shows the platelet aggregates observed in the well using the Impact-R camera and scan electron microscopy (SEM). Using SEM, the platelet spreading is clearly observable on small aggregates.

 

Fig. 1 (a) Platelet aggregates picture obtained with the Impact-R camera (original magnification x400). The whole blood was exposed to a shear rate of 100 s⁻¹. (b) Scan Electron Microscopy of platelet aggregates in the well.

Download Full Size | PDF

DHM setup

This section describes the off-axis DHM with a source of partially spatial coherence light to record the holographic information [14]. The configuration is shown in Fig. 2.

 

Fig. 2 Digital holographic microscope schema: GG: Rotating Ground Glass; L1 and L2: Lenses; BS1 and BS2: Beam Splitters; ML1-3: Microscope Lenses, M1-5: Mirrors.

Download Full Size | PDF

A coherent source (a mono-mode laser diode, λ = 532nm) is made partially spatial coherent by focusing the beam, by the lens ML1, close to the rotating plane of the ground glass (GG). The lens L1 collimates the beam that is divided by a beam splitter BS1. The object beam reflected by BS1 illuminates the sample in the analysis chamber in transmission. The plane, on which the platelet aggregates are sticking, is imaged by the couple of lenses ML3-L2 on the CCD camera sensor. The reference beam transmitted by the beam splitter BS1 has a similar optical path, excepted that there is no analysis chamber with a sample. The two beams are interfering on the CCD sensor. The reference beam is slanted on the sensor with respect to the object beam in such a way that a grating-like thin interference pattern is recorded. This off-axis configuration enables to implement the Fourier method to compute the complex amplitude of the object beam for every recorded frame [18]. Thanks to the partially coherent illumination, raw holograms directly displayed on the PC screen are fully meaningful for the operator as with a usual microscope (Fig. 4(b)). With full coherent illumination, the direct image displayed on the screen is usually too noisy to be interpreted by a direct viewing.

The microscope lenses ML2 and ML3 are Leica 40x, NA 0.6. The camera is a JAI CV-M4, with a CCD array of 1280 x 1024 pixels cropped at 1024 x 1024 pixels, with a pixel size of 7.4µm x 7.4µm. The field of view is 165 µm x 165 µm

Holograms acquisition and background correction

As observed in Fig. 3(a), the recorded fields of view of the DHM show aggregates disseminated on a background field. The main objective is to measure the aggregate shapes thanks to the quantitative phase contrast imaging capability of the DHM. In this section, we describe the processing steps to obtain the complex amplitude information and the resulting phase information on the aggregates.

 

Fig. 3 (a) Phase image showing platelet aggregates obtained with DHM (scale bar = 20µm). The average phase value is arbitrarily set to be equal to the grey level 30. (b) The phase image of (a) is multiplied by a binary mask in order to keep the sole background region. For visibility purpose, the phase values are multiplied by a factor 2.

Download Full Size | PDF

In digital holographic microscopy, optic elements can introduce minor distortions in the background phase. This is particularly pertinent when the main parameter analyzed is the phase information that allows to quantify the morphological shapes of the platelet aggregates. For that purpose, it is necessary to implement a phase background subtraction and permanent defects elimination. The phase background subtraction is directly implemented in the recorded plane, or equivalently, in the imaged plane, since the images of the aggregates are focused and there is no need to perform digital holographic refocusing. However, the sample already included the aggregates, making it difficult to perform the background phase subtraction on the basis of a single hologram. In order to overcome this issue, we first recorded a sequence of N holograms (in the present article, N = 29) of the sample moved with lateral translations. This sequence of holograms allows us to implement corrections of the defects in the intensities and the phase maps in a self-consistent way. The defects correction uses the procedures that are described as follow.

The S complex amplitude$g_k\left(s,t\right)$, where $\left(s,t\right)$ are the discrete spatial variables, with $k=0,\mathrm{...},N-1$, $s,t=0,\mathrm{...},n-1$ and n is the pixel number by side, are extracted from a set of recorded hologram $h_k\left(s,t\right)$. The original holograms have a size of 2048x2048 pixels, while the complex amplitudes are reduced to 1048x1048 pixels.

A first step of correction on the intensity field is performed. For that purpose, the averaged intensity $i_a\left(s,t\right)$ of the intensities $i_k\left(s,t\right)={\left|g_k\left(s,t\right)\right|}^2$ is computed. The corrected intensities are computed thanks to:

For the correction of the phase maps ${\phi }_k\left(s,t\right)$ associated to$g_k\left(s,t\right)$, the averaged phase map ${\phi }_a\left(s,t\right)$is computed and subtracted to every phase map ${\phi }_k\left(s,t\right)$ to obtain the corrected phase maps ${\phi }_{ck}\left(s,t\right)$ according to:

The phase background is after set, in average, to a fixed phase value. In our case, for the phase values ranging on 255 levels (1 byte), we set the background to the level 30 to avoid phase jumps. The process is efficient and is illustrated by the Fig. 3(b).

On each recorded hologram, the physical heights $h_k\left(s,t\right)$ are determined by:

Statistics

Sigma Plot 12.0 was used for the statistical analysis. Univariate analysis was depicted by Pearson’s coefficient. Comparison between the aggregates parameters, obtained with the whole blood exposed to a shear rate of 100 s⁻¹, after 20 sec and 300 sec, were done using Mann-Whitney Rank Sum. Differences were considered as statistically significant with a two-tailed p<0.05 (reject the null hypothesis, H₀)

3. Results and discussion

Platelet aggregate detection

In order to determine the phase error, the regions covered by the aggregates were detected to evaluate the phase fluctuations outside these regions. The recorded fluctuations are then used to establish the accuracy of the phase measurements. The detection processing that we used is identical to the one described in reference [14]. It is based on a high-pass filtering of the $g_{ck}^{}\left(s,t\right)$. Indeed, when there is locally no aggregate in some areas of the field of view, $g_{ck}^{}\left(s,t\right)$ is almost constant in this region and the high-pass filtering process gives complex amplitude with very low module values that are eliminated by a simple threshold operation. On the contrary, the high-pass filtering enhances the presence of an aggregate by local strong complex amplitude. The high-pass filtering is used to detect the aggregate regions in order to realize a set of binary masks covering those regions. In the same time, the inverted masks give the regions where there is no aggregate (background). The regional information is then used to measure the phase background fluctuations and the phase heights of the aggregates on the phase images from the unfiltered amplitudes.

To avoid border effects on the high-pass filter, we implemented, as in reference [14], an inverse Gaussian filter $H\left(u,v\right)$ defined by:

where $\left(u,v\right)$are the discrete spatial frequencies ($u,v=-n/2,\mathrm{...},n/2-1$), and $\sigma$is the width of the high-pass filter. In practice, $\sigma =10$gives good results. After the inverse Fourier transformation in the filtering process, a threshold is applied. The intensity image is computed and converted into 255 gray levels in such a way that aggregates give rise to bright regions, even with saturation, and that the background regions give intensity of few grey levels (typically less than 10). The threshold level we applied is 40. It results binary images constituted by a dark background in which there are unconnected bright regions. A surface analysis is performed in order to eliminate the smaller areas (<4 pixels). A morphological dilatation is performed with a 10-pixels diameter structural element in order to guarantee that the bright areas partly cover already the background region.

As we want to assess the fluctuation of the phase background, we invert the contrast to obtain a mask on which the background fluctuations are computed. An example of result is shown in Fig. 3.

The full set of phase images is used to assess the fluctuations of the phase background by computing the standard deviation that is given by StDev = 3.2 grey level. Considering this value as the typical error on the phase and tanks to Eq. (3), we obtain, that the error $\Delta h$ on the height of the aggregates is 17.05nm. By using classical statistical tools, the error on the volume of an aggregate $\Delta V$can be expressed by:

where s is the pixel area and N the number of pixel covered by the aggregate. Thanks to Eq. (5), assessment of the error on the volume computation is calculated as follow.

Computation of the volume of the platelet aggregates

The volume of the aggregate a in the hologram k is obtained by computing in the corresponding phase image:

1/ a morphological dilatation on each non-zero valued zone of the binary detection image obtained by the method described above.

2/ as each resulting zone will exceed the actual area of the corresponding aggregate, a low-level threshold operation is performed (Threshold value = background level + 3 grey levels) in order to keep the only pixels belonging to the aggregate.

However, with this method it is not excluded to have overlaps between some neighbor aggregates that could influence the assessment of the aggregate volume. For that reason, we decided to select regions of interest of aggregates that are unambiguously surrounded by the background to avoid the overlaps effect. This was performed to have a statistical relevance. On the basis of 29 original holograms corresponding to shear rate exposition times of 20s and 300s, we processed a total of 340 platelet aggregates. To observe the platelet aggregates spreading process, we exposed whole blood to a constant shear rate (100s⁻¹) during 20 sec and 300 sec. As first indications, for those expositions, the averaged platelet volumes are, 4.50 ± 0.015µm³ and 3.74 ± 0.015µm³ respectively. More information was extracted as described below.

The Fig. 4 summarizes the different steps of the analysis. The well of the impact-R test is placed on the motorized table of the DHM for analysis (Fig. 4(a)). The Fig. 4(d) shows a zoom on an aggregate and the 3D representation extracted from the phase, where the spreading is clearly observable.

 

Fig. 4 (a) The well positioned on the DHM for analysis, (b) Hologram with a zoom on platelets aggregate. (c) Hologram intensity d) Phase image and in the box, the phase representation with the 3D extraction of a platelets aggregate based on the optical high (grey scale). (scale bar = 20µm)

Download Full Size | PDF

DHM enables the calculation of the maximal heights, surfaces and volumes of platelets and aggregates. In Fig. 5, we report the heights, surfaces and volumes of aggregates obtained with the whole blood exposed to a shear rate of 100 s⁻¹, after 20 sec and 300 sec.

 

Fig. 5 (a), (b), (c) Comparison between heights, surfaces and volumes of platelets aggregates obtained with the whole blood exposed to a shear rate of 100 s⁻¹, after 20 sec and 300 sec. (d) and (e) Correlations between the aggregates volumes and the surface-height product obtained at 20 sec and 300 respectively (more than 160 aggregates were analyzed in each condition).

Download Full Size | PDF

Each aggregate has an area represented by the number of pixel area covered by the aggregate and a volume obtained by the sum of the individual volume on each pixel. For a largely spread aggregate, its area can be large while the volume on each pixel is low. On the contrary, in case of weakly spread aggregate, its area can be small while the volume for each pixel is high. The Figs. 5(d) and 5(e) are representative of the aggregate spreading. A significant decrease is observed after 300 sec for the heights (p<0.001) and the volumes (p = 0.003). In contrast, the surfaces increase after 300 sec (p<0.001). On the Figs. 5(d) and 5(e), we plot the correlations between the volumes (Vol) and the product of the maximum height by the aggregate surface (S x H). The volume Vol of several simple 3D objects can be expressed as Vol = a*S*h where S is the area of the base and h the height of the object. The coefficient “a” depend on the geometry of the object: a = 1 for a cylinder and a = 1/3 for a cone. Figures 5(d) and 5(e) indicate that a = 0.24 at t = 20s and a = 0.2 at t = 300. This suggests that the aggregates are thinner than cones. In other words, the changes of the angular coefficients at 20 sec and 300 sec express the spreading effect. It is the first time that these values can be calculated.

Platelet receptors and cytoplasmic molecules, such as calpain-1 [20] and talin [21] which are involved in signaling cascades after adhesion and activation have been extensively studied. However, the effect of these cascades on plug morphology have been commonly analyzed through the transmission microscopy alone or combined with fluorescence in 2 dimensions only [11,20,22,23]. The DHM is very convenient to use and gives more information on the effect of the molecules involved in the spreading on the nascent aggregate shape. By using a flow chamber within the DHM, we plan to perform a kinetic study of platelet aggregates evolution in 3D in different clinical situations.

4. Conclusion

Here, we report for the first time analysis of platelets aggregates by DHM. This technique enabled the study the morphology of the platelets adhesion, aggregation and spreading in in vitro models. This original method is particularly well-suited for the study of the platelets physiology, the physiopathology in clinical practice and the development of new drugs.